Multiple direct and indirect roles of the Paf1 complex in transcription elongation, splicing, and histone modifications

Abstract

The polymerase-associated factor 1 (Paf1) complex (Paf1C) is a conserved protein complex with critical functions during eukaryotic transcription. Previous studies showed that Paf1C is multi-functional, controlling specific aspects of transcription ranging from RNA polymerase II (RNAPII) processivity to histone modifications. However, it is unclear how specific Paf1C subunits directly impact transcription and coupled processes. We have compared conditional depletion to steady-state deletion for each Paf1C subunit to determine the direct and indirect contributions to gene expression in Saccharomyces cerevisiae. Using nascent transcript sequencing, RNAPII profiling, and modeling of transcription elongation dynamics, we have demonstrated direct effects of Paf1C subunits on RNAPII processivity and elongation rate and indirect effects on transcript splicing and repression of antisense transcripts. Further, our results suggest that the direct transcriptional effects of Paf1C cannot be readily assigned to any particular histone modification. This work comprehensively analyzes both the immediate and the extended roles of each Paf1C subunit in transcription elongation and transcript regulation.

Keywords: CP: Molecular biology; Cdc73; Ctr9; H3K36me3; Leo1; Paf1; Paf1 complex; RNA polymerase II; Rtf1; histone modifications; transcription elongation.

Figure 1.. Deletion of genes encoding Paf1C subunits causes subunit-specific effects on nascent and steady-state transcriptomes and RNA stability

(A) Scheme to profile total and nascent transcriptomes in Paf1CΔ mutants. (B and D) Heatmaps depicting log₂-fold change from −500 to +5,000 bp relative to the transcription start site (TSS) in 4tU-labeled (B) and total RNA (D) at genes sorted by length (n = 4,232).^, Data end at the CPS. (C and E) MA plots of differential gene expression over protein-coding sequences (n = 6,600) calculated from 4tU-seq (C) or total RNA-seq data (E). (F) Relative stability of transcripts calculated as log₂-fold change in the ratio of total to nascent RNA relative to WT. (G) Gene ontology analysis of the change in transcript stability for Paf1CΔ mutants. Above 0 indicates ontology categories for which RNAs are more stabilized than expected in a mutant (overUnder = 1), and below 0 indicates classes less stable than expected (overUnder = −1). Color indicates mutant condition as in (A). Dashes indicate false discovery rate (FDR) = 0.05. Significance was determined by Mann-Whitney U test of WT and Paf1CΔ data: *p < 0.05, **p < 0.01, and ***p < 0.001. See also Figure S1.

Figure 2.. Rapid depletion of Paf1C subunits leads to widespread downregulation of transcript synthesis

(A) Scheme to profile nascent transcriptomes in Paf1C-AID strains. Rainbow coloring is meant to indicate that each Paf1C subunit was depleted individually. (B) Western analysis of strains treated for 60 min with DMSO (Veh) or depleted of Paf1C subunits by treatment with auxin for 0 min (Pre) or 5–60 min. (C) Heatmaps depicting log₂-fold change in spike-in normalized 4tU-seq reads for cells depleted of the indicated Paf1C subunit relative to undepleted conditions (Veh) for 30 min. Genes are sorted by length. (D and E) MA plots showing differential expression of SRATs in Paf1CΔ strains (D) or auxin-treated Paf1C-AID strains (E) relative to respective control, WT or Veh. See also Figure S2.

Figure 3.. Paf1C directly impacts RNAPII processivity

(A) Calculation of CS. (B) Cumulative distribution of log₂-fold change in 4tU-seq CS in Paf1CΔ strains or Paf1C-AID + auxin conditions relative to respective control (WT or Veh). (C and E) Log₂-fold change in 4tU-seq CS stratified by gene length. (D and F) Browser tracks depicting sense-strand 4tU-seq signal over the 12.3 kb DYN1 gene. (G) Log₂-fold change in CS as in (A) from α-Rpb3–3xFLAG (RNAPII) ChIP-seq data. (H) Heatmaps of the log₂-fold change in RNAPII occupancy in ctr9Δ or Ctr9-depleted conditions relative to WT or Veh control on protein-coding genes. Significance was determined by Mann-Whitney U test of control and experimental data with FDR correction: *p < 0.05, **p < 0.01, and ***p < 0.001. See also Figure S3.

Figure 4.. Simulations describe altered distribution of RNAPII elongation rate and processivity across gene bodies

(A) RNA synthesis per base pair (left) and RNAPII distribution (RNAPII/bp) over gene bodies (right) are inferred with techniques that empirically derive imperfect quantitation via 4tU-seq and RNAPII ChIP-seq. Simulations are initialized with a set of input parameters, including locus length (L; bp), processivity (PC), elongation rate (ER; bp/min), background signal, labeling time, and an initial RNAPII flux (RNAPII/min). Elongation dynamics simulations produce idealized 4tU-seq and ChIP-seq profiles that must simultaneously match nascent transcriptomic and RNAPII profiling data. (B) Diagram depicting zones of transcription in simulations. Adjustable parameters are in purple. Population-level simulations of loci with variable lengths are initialized with a prescribed frequency of initiation (Flux_init), 4tU-RNA labeling time, and background signal from RNAPII ChIP-seq, which were kept constant across all conditions (set to 1 RNAPII/min, 5 min, and 0.014 arbitrary units, respectively). Each zone has an adjustable length (L_n) except for L₂, which varies depending on gene length. RNAPII-occupied zones 1–4 can be provided adjustable elongation rate (ER_n) or processivity (PC_n) parameters, which linearly ramp between zones as diagrammed. Elongation and processivity parameters in zone 2 were kept constant between the end of zone 1 and the start of zone 3. (C–F) Comparisons of simulations (Sim) and averaged empirical data normalized by the level of genic signal excluding the top and bottom 5% of normalized signal at each bin. Data and simulations for two gene length classes are shown. (C and D) 4tU-seq (C) and RNAPII ChIP-seq (D) comparisons of Ctr9-AID undepleted (Veh) or depleted (Aux) conditions. (E and F) 4tU-seq (E) and RNAPII ChIP-seq (F) comparisons of WT and ctr9Δ conditions. Shaded areas represent standard deviation of empirical data. (G) Parameter values passed into simulations displayed in (C)–(F). Scale bars indicate change relative to respective control conditions. Gray fill indicates parameters are unchanged across all simulations. See also Figures S4 and S5.

Figure 5.. Long-term absence of Paf1C subunits leads to splicing defects

(A and B) Browser tracks showing 4tU-seq read density over RPS14B. Dotted line denotes intron. (C) Diagram depicting the calculation of fraction of unspliced reads over the 5′ splice junction. (D and E) Violin plots of log₂-fold changes in the fraction of unspliced reads at 5′ splice junctions for Paf1CΔ (D) or Paf1C-AID strains treated with auxin (30 min) (E). (F) Heatmap showing log₂-fold change in fraction of unspliced reads over 5′ splice sites in ribosomal protein (RP) and non-RP genes sorted by fold effect of paf1Δ condition. (G and H) Metaplots of RNAPII-Ser2P ChIP-seq signal over protein coding genes. Significance determined by Mann-Whitney U test of control and experimental data: *p < 0.05, **p < 0.01, and ***p < 0.001. See also Figure S6.

Figure 6.. Loss of H3K36me only partially explains phenotypes of paf1Δ strain

(A) Western analysis of H3K36me3 and H3K36me2 levels. (B and C) Quantitation of H3K36me signals in (A) plus two additional biological replicates relative to G6PDH. Normalized WT values were set to 1. (D and E) Log₂-fold change in 4tU-seq reads of SRATs in indicated mutants relative to WT presented as metaplots (D) or heatmaps depicting individual loci (E). Start and end refer to annotated 5′ and 3′ ends of SRATs sorted by mean fold change across all conditions. (F) Log₂-fold change in the fraction of unspliced 4tU-seq reads relative to WT. (G) Log₂-fold change in CS as calculated in Figure 3A from 4tU-seq data, relative to WT. Significance for western blot data was determined by two-tailed t test in comparison to WT. Significance for other data was determined by Mann-Whitney U test of control and experimental data with FDR correction for CS measurements: *p < 0.05, **p < 0.01, and ***p < 0.001. See also Figure S7.

Figure 7.. Evolving consequences of the loss of Paf1C

Model depicting changes to transcriptional behavior, RNA splicing, and chromatin from WT state to states of acute Paf1C subunit depletion or Paf1C subunit deletion.